Problem with the CONVERT* signal (PCI-6110 board).

[remifab]

I am using Labview 6.1 / Window 2000 and a PCI�6110 card.
My system uses an external clock (connected to the PFI7) to sample a signal in channel 0. The program controlling the DAQ is a modified version of �Acquire N scans ExtScanClk D-Trig,vi� example, I have removed from this example the two sub-vi �AI Clock Config� as the PCI-6110 sample each channel simultaneously. I have also had the �Route Signal.vi� to route the CONVERT* to the PFI2.

To test my system I used a signal generator at 1MHz as EXTERNAL clock and a 50kHz sinewave signal as entry in channel 0. If I do a Fourier transform of 1mega samples of the acquired signal I obtained a peak centred at 50kHz plus several minor sidebands (as obtained with a frequen
cy or phase modulation). With a 8Gs/s digital oscilloscope I found that my generator signal is stable but the CONVERT* signal have a variable width in time: the signal rises after a fixed time after the falling edge of the clock but falls between 80 and 100ns after.

If I use the INTERNAL clock of the board (set at 1MHz), I obtained the previous peak at 50kHz but without the sidebands. The CONVERT* signal is stable in time (always ~80ns).

According to the PCI-6110, a high to low edge on CONVERT* indicates that an A/D conversion is occurring. So if I use an EXTERNAL clock, there is an uncertainty of 20ns on the time when an A/D conversion happens.

My questions:

1) As I am pretty sure that the sideband comes from this variable width of CONVERT*, does a way exit to make sure that the A/D conversion always occurs after a fixed delay from the falling edge of my external clock?
2) Is it normal that the board works like this?

Thank you for your answer.

[RIO-Todd]

I have never heard of this happening when using an external clock with a device. What specific VI are you using to generate the FFT of the signal? Are you using a VI that accepts the waveform input? The reason I'm asking is that if you're using an external scan clock and passing the waveform from the AI Read VI to an FFT VI, you won't be passing the time information along with the waveform type. Therefore, you will want to make sure you change the waveform cluster to include the delta t of 0.000001 sec (for a 1 MHz external clock). This will make sure that the Fourier Transform VI is using the correct sampling information when converting to the frequency domain.

I hope this helps.

Regards,

Todd D.
NI Applications Engineer

[remifab]

Thank you for your help,

Actually the problem doesn�t come from a FFT.vi as I am using Matlab to produce my FFT (the file transmitted to Matlab contains only the data amplitude) and I correctly give the good parameters (e.g the correct sampling information).

Using Matlab, the correct sampling information, an INTERNAL clock: good result and the CONVERT* signal is stable in time (always ~80ns).

Using Matlab, the correct sampling information, an EXTERNAL clock: bad result and CONVERT* signal have a variable width in time: the signal rises after a fixed time after the falling edge of the clock but falls between 80 and 100ns after.

So the problem comes from the CONVERT* signal.
Why this signal is not constant in time using an external clock?
Is
the falling edge of this signal synchronized with the internal base clock at 20MHz.?
If so does a way exit to make sure that the A/D conversion always occurs after a fixed delay from the falling edge of my external clock?

Regards

[RIO-Todd]

Sorry for the delay. I programmed a 6110 to use an external clock and looked at the convert signal just as you are doing. I do see the fluctuation of the convert signal when using an external clock.

I also am performing an FFT within LabVIEW, but I don't see a difference in the FFT when looking at the date with and without an external clock. Is your external clock precise? I have made sure my external clock is stable and do not see the difference in FFTs but have seen the difference in the convert signal.

Regards,

Todd D.

[remifab]

Todd D,
Thank you for your investigation. I am sure that my clock is stable.
So the difference must be with the FFT. With Labview I record only the value of the data (not the time when the data are recorded). I assume in both cases (internal and external clock) the samples are evenly taken (e.g. the time between two samples is 1/1MHz = 1us) when I am performing a FFT. This is true with the internal clock but not totally true with the external, as there is a fluctuation of the convert signal.
I reckon that with you program Labview �knows� the real time when the samples have been acquired and they are not evenly spaced.
Which Vi are you using and do you give to the Vi the real time when the samples are taken?
So it is possible to get t
he time when the sample at taken?
Todd, are you able to send me the code of you program?

Thank you again for the time you spent to my problem.
Regards,

RemiFab

[RIO-Todd]

Hey RemiFab,

I've attached the VI that I used to get the FFT of the signal in LabVIEW. The VI uses the 'FFT Power Spectrum VI' to compute the FFT. I hope this helps.

Regards,

Todd